WHEN VERIFICATION RULES CHANGE Adversarial Evidence, Cross-Revision Warrant, and Exploratory Negative Controls for AAC

Scientific claims about physical transformations depend not only on the reliability of observational records but also on the conditions under which those records retain evidential authority when verification rules change. This paper examines that problem through adversarial epistemology, scientific measurement, and the critical reconstruction of Naoto Fujie’s Operational-Shape Realism (OSR). The argument begins with the possibility that physically distinct causal histories generate observationally equivalent records. Additional measurements may restore discrimination, but shared calibration dependencies, compromised provenance, and revisions to validation procedures introduce further epistemic constraints. These difficulties motivate an examination of cross-revision warrant: the conditions under which evidence acquired under one verification regime may legitimately support claims under another. A restricted statistical result establishes a bound on changes in discriminability when the relevant evidence distributions remain close in total variation distance. This result does not independently establish causal identification or epistemic justification. The paper also examines the Adversarial Attribution Constraint (AAC) through two exploratory synthetic studies involving 1,000 and 1,200 scenarios. The results reveal no independent advantage over a comparator equipped with equivalent information and decision rules. They consequently support a restrictive interpretation of AAC as an integrative auditing procedure rather than an established novel philosophical principle. Finally, the paper revisits its criticism of OSR. Since Fujie’s accessible research manuscripts already distinguish real transformations from warranted epistemic attribution, the initial objection that OSR conflates reality with observation is withdrawn. The remaining questions concern the evidential conditions of particular scientific attributions rather than a demonstrated contradiction in OSR. The paper develops a conditional account of evidential transfer across verification regimes while identifying the limits of its exploratory computational evidence.